Rolling bearing fault time-varying instantaneous feature enhancement and extraction method

Abstract

The invention discloses a rolling bearing fault time-varying instantaneous feature enhancement and extraction method. The method comprises the following steps: acquiring an original bearing fault signal; decomposing an original bearing fault signal based on an adaptive noise complete set empirical mode decomposition method to obtain a series of intrinsic mode function components; component properties are distinguished based on a correlation coefficient hopping criterion, and denoising is carried out; mapping the bearing fault signal after noise reduction to a time-frequency chirp rate space based on a local maximum frequency chirp rate synchronous compression chirp transformation method; and extracting a time-frequency chirp rate coefficient reflecting fault time-varying instantaneous characteristics from the extruded time-frequency chirp rate space, reconstructing a bearing fault impact component, performing Hilbert envelope analysis on the bearing fault impact component, and identifying the characteristic frequency of the bearing fault. According to the method, the fault time-varying instantaneous characteristics of the rolling bearing can be clearly and accurately extracted, and the method has a good engineering application prospect.

Description

Technical Field

[0001] The invention belongs to the technical field of bearing fault feature extraction, and in particular relates to a method for enhancing and extracting time-varying transient features of rolling bearing faults. Background Art

[0002] As a key component in rotating mechanical equipment, the operating status of rolling bearings directly affects the safety of the entire equipment. Testing and analyzing the vibration signal of rolling bearings is currently an important means to determine the working status of bearings. The vibration signal of rolling bearings is a typical multi-component non-stationary signal, and its time-varying transient characteristics just reflect the state characteristics of the bearing. The time-frequency analysis method is widely used in the analysis of bearing vibration signals because it can obtain the time-varying transient characteristics of the signal. However, for early-stage rolling bearing failures, the time-varying transient characteristics of the vibration signal are weak and are interfered by noise and transmission paths. The effect of time-frequency analysis and the identification of time-varying transient characteristics will be affected, thereby increasing the difficulty of fault feature extraction.

[0003] At present, the original bearing vibration signal is decomposed by wavelet or wavelet packet decomposition, empirical mode decomposition (EMD), local mean decomposition, singular value decomposition and other non-stationary nonlinear decomposition methods to remove noise components and retain characteristic components to reconstruct the time spectrum, which is an important bearing vibration signal denoising method. However, the effect of wavelet or wavelet packet decomposition depends largely on the wavelet basis function and decomposition layer number determined before decomposition, and has poor adaptability; decomposition methods such as EMD are adaptive, but it is difficult to accurately distinguish the noise component and signal component for the early fault vibration signal of the bearing with low signal-to-noise ratio; and the denoising effect of a single decomposition is generally limited, and a certain amount of noise components may still remain.

[0004] In terms of enhancing faulty time-varying instantaneous features, time-frequency energy aggregation and time-frequency ridge extraction methods such as Synchro Squeezing Transform (SST) and its variants, time-scale manifold ridge demodulation, and dynamic path optimization can gather energy near time-varying instantaneous feature components, or obtain time-frequency ridges that reflect time-varying instantaneous features, that is, enhance or clarify the corresponding time-varying instantaneous features. At present, they have been applied to the extraction of faulty time-varying instantaneous features of bearing vibration signals. However, for early bearing fault signals, due to the influence of noise and unclear time-varying feature components, simple SST or time-frequency ridge extraction may result in compression errors or unconcentrated compression energy, as well as errors in or inability to extract time-frequency ridges.

[0005] In view of the above problems, it is urgent to propose a method for enhancing and extracting the time-varying transient features of rolling bearing faults. Summary of the invention

[0006] In order to solve the above technical problems, the present invention proposes a method for enhancing and extracting the time-varying transient characteristics of rolling bearing faults to solve the problems existing in the above prior art.

[0007] To achieve the above object, the present invention provides a method for enhancing and extracting time-varying transient features of rolling bearing faults, comprising the following steps:

[0008] Collect original bearing fault signals;

[0009] Decomposing the original bearing fault signal based on an adaptive noise complete set empirical mode decomposition method to obtain a series of intrinsic mode function components;

[0010] The properties of the intrinsic mode function components are distinguished based on the correlation coefficient jump criterion, and denoising is performed by a secondary decomposition and secondary reconstruction method to obtain a bearing fault signal after denoising;

[0011] Based on the local maximum frequency chirp rate synchronous compression chirp transform method, the de-noised bearing fault signal is mapped to the time-frequency chirp rate space through chirp transform;

[0012] The time-frequency chirp rate coefficient reflecting the time-varying instantaneous characteristics of the fault is extracted from the squeezed time-frequency chirp rate space, the bearing fault impact component is reconstructed, the bearing fault impact component is subjected to Hilbert envelope analysis, and the characteristic frequency of the bearing fault is identified.

[0013] Optionally, the process of decomposing the original bearing fault signal based on an adaptive noise complete set empirical mode decomposition method to obtain a series of intrinsic mode function components includes:

[0014] The original bearing fault signal is taken as the signal to be processed, and different Gaussian white noises are added to the signal to be processed to obtain a noisy signal set; each noisy signal in the noisy signal set is decomposed based on the empirical mode decomposition method, and the first-order intrinsic mode function component is obtained by ensemble averaging; the first-order intrinsic mode function component is removed from the signal to be processed to obtain a first residual signal; the empirical mode decomposition is continued on the first residual signal after noisy, and the second-order intrinsic mode function component is obtained by ensemble averaging; the second-order intrinsic mode function component is removed from the first residual signal to obtain a second residual signal; and so on, until the obtained residual signal is a monotonic function, the iteration is terminated, and finally a series of intrinsic mode function components are obtained.

[0015] Optionally, the process of distinguishing the properties of the intrinsic mode function components based on the correlation coefficient jump criterion includes:

[0016] The correlation coefficient jump criterion adopts the Pearson correlation coefficient to obtain the correlation coefficient value between the vectors, obtains the maximum jump value of the correlation coefficient of the intrinsic mode function component based on the correlation coefficient value between the vectors, determines the maximum jump value of the correlation coefficient of the intrinsic mode function component, and distinguishes the noise component and the non-noise component in the intrinsic mode function component.

[0017] Optionally, the process of mapping the de-noised bearing fault signal to a time-frequency chirp rate space through chirp transform based on the local maximum frequency chirp rate synchronous compression chirp transform method includes:

[0018] The first theorem, the second theorem and the local maximum FC estimator are constructed. Based on the first theorem, the second theorem and the local maximum FC estimator, the denoised bearing fault signal is mapped to an independent region of the time-frequency chirp rate space through chirp transform.

[0019] Optionally, the formula of the first theorem is as follows:

[0020] Define a set of non-empty sets Ψ and Ψ h as follows:

[0021]

[0022] in, A h (t) and φ h (t) are s h (t) is the instantaneous amplitude and instantaneous phase, C s (t, μ, c) is the CT transform of s(t) based on the Gaussian window function, μ is the instantaneous frequency, and c is the chirp rate;

[0023] When s(t) satisfies the following amplitude conditions:

[0024]

[0025] In the formula, Λ1=(Δ 2 +1) -1 / 4 , Λ2=(4Δ 2 +1) -1 / 4 , Δ is the separation resolution; α>0;

[0026] Ψ is represented by Ψ h The non-intersecting form of is:

[0027]

[0028] Optionally, the formula of the second theorem is expressed as follows:

[0029] When s(t) satisfies the above amplitude conditions, the following results hold:

[0030]

[0031] In the formula, is the weight s h (t) is an estimate of the instantaneous frequency and chirp rate; φ' h (t),φ″ h (t) are the components s h Theoretical values ​​of the instantaneous frequency and chirp rate of (t), A l (t) is the component s l (t), the instantaneous amplitude when l≠h.

[0032] Alternatively, the formula of the local maximum FC estimator is as follows:

[0033]

[0034] In the formula, Δ μ , Δ c They are frequency resolution and chirp rate resolution respectively.

[0035] The present invention also provides a computer device, comprising a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method.

[0036] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the method are implemented.

[0037] The present invention also provides a computer program product, comprising a computer program, which implements the steps of the method when executed by a processor.

[0038] Compared with the prior art, the present invention has the following advantages and technical effects:

[0039] At present, the working environment of most rolling bearings is characterized by large background noise, poor working conditions, and complex system structures. It is difficult to detect weak faults in the early stage, which can easily lead to serious faults and main engine shutdown, resulting in serious economic losses and safety accidents. Compared with the existing methods, the present invention proposes a method for enhancing and extracting the time-varying transient characteristics of rolling bearing faults. Firstly, the original bearing fault signal is decomposed based on the adaptive noise complete set empirical mode decomposition method to obtain a series of intrinsic mode function components; the properties of the intrinsic mode function components are distinguished based on the correlation coefficient jump criterion, and the components are removed by secondary decomposition and secondary reconstruction. The noise is removed to obtain the bearing fault signal after noise reduction; the bearing fault signal after noise reduction is mapped to the time-frequency chirp rate space through chirp transform based on the local maximum frequency chirp rate synchronous compression chirp transform method; and after squeezing and rearranging, the energy can be gathered near the natural frequency of the bearing, thereby enhancing and highlighting the time-varying instantaneous characteristics characterizing the fault, while removing the interference of sidebands and residual noise; then the time-frequency chirp rate coefficient reflecting the time-varying instantaneous characteristics of the fault is extracted from the squeezed time-frequency chirp rate space, and the bearing fault impact component is reconstructed, and the Hilbert envelope analysis is performed on the bearing fault impact component to identify the fault characteristic frequency of the bearing.

[0040] The present invention comprehensively considers the removal of non-fault characteristic components such as noise, modulation side frequencies, rotation frequencies, and the enhancement of fault characteristics. It is particularly suitable for early fault vibration signals of rolling bearings with weak fault characteristics, severe noise interference, and complex modulation side frequency bands. It can more clearly and accurately extract the faulty time-varying instantaneous characteristics of rolling bearings, improve the recognition rate of early bearing faults, ensure the safety and reliability of the equipment, and has good engineering application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0042] Figure 1 is a flow chart of a method according to an embodiment of the present invention;

[0043] Figure 2 A schematic diagram of a rolling bearing inner ring fault simulation signal according to an embodiment of the present invention;

[0044] Figure 3 Schematic diagram of a CEEMDAN decomposition result of an embodiment of the present invention, wherein (a) is a schematic diagram of the decomposition result of the original signal and IMF1,1 to IMF1,7, and (b) is a schematic diagram of the decomposition result of IMF1,8 to IMF1,14 and the residual;

[0045] Figure 4A CEEMDAN decomposition result correlation coefficient curve diagram of an embodiment of the present invention;

[0046] Figure 5 A time domain waveform diagram of a simulated signal reconstructed once according to an embodiment of the present invention;

[0047] Figure 6 A correlation coefficient curve diagram of the secondary CEEMDAN decomposition result of an embodiment of the present invention;

[0048] Figure 7 Schematic diagram of a secondary reconstructed signal and its Hilbert envelope spectrum according to an embodiment of the present invention, wherein (a) is a schematic diagram of a time domain waveform, and (b) is a schematic diagram of a Hilbert envelope spectrum;

[0049] Figure 8 is a STFT time-frequency diagram of the secondary reconstructed signal according to an embodiment of the present invention;

[0050] Fig. 9 It is a time-frequency diagram after ELMSSCT processing in an embodiment of the present invention;

[0051] Fig.10 Schematic diagram of the reconstructed bearing fault signal and its Hilbert envelope spectrum according to an embodiment of the present invention, wherein (a) is a schematic diagram of the time domain waveform and (b) is a schematic diagram of the Hilbert envelope spectrum. DETAILED DESCRIPTION

[0052] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0053] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0054] Embodiment 1

[0055] like Figure 1As shown, in this embodiment, a method for enhancing and extracting time-varying transient features of rolling bearing faults is provided. The method firstly innovatively establishes a correlation coefficient jump criterion (CCJC) according to the characteristics of the bearing vibration signal to distinguish the noise component and non-noise components such as periodic fault impact, rotation frequency and its multiple frequency in the bearing vibration signal; then, the non-stationary nonlinear data processing capability with good regularity of Complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) is utilized to fully reduce the noise of the original bearing vibration signal through secondary decomposition and secondary reconstruction; on this basis, the local maximum frequency chirp rate synchronous compressed chirp transform (Enhanced local maximum frequency-chirp-rate synchrosqueezedchirplet The denoised bearing signal is mapped to the time-frequency-chirp-rate (TFC) space through the chirp transform (ELMSSCT) method, and the energy of each characteristic component in the signal is concentrated by squeezing and rearranging the frequency-chirp rate coefficients in the TFC space, thereby enhancing the time-varying instantaneous characteristics reflecting the bearing fault and improving the distinguishability of similar components of the instantaneous characteristics, laying a good foundation for finally obtaining clear and clean bearing fault feature information.

[0056] As an implementable manner, the method comprises the following steps:

[0057] Collect original bearing fault signals;

[0058] Decomposing the original bearing fault signal based on an adaptive noise complete set empirical mode decomposition method to obtain a series of intrinsic mode function components;

[0059] The properties of the intrinsic mode function components are distinguished based on the correlation coefficient jump criterion, and denoising is performed by a secondary decomposition and secondary reconstruction method to obtain a bearing fault signal after denoising;

[0060] Based on the local maximum frequency chirp rate synchronous compression chirp transform method, the de-noised bearing fault signal is mapped to the time-frequency chirp rate space through chirp transform;

[0061] The time-frequency chirp rate coefficient reflecting the time-varying instantaneous characteristics of the fault is extracted from the squeezed time-frequency chirp rate space, the bearing fault impact component is reconstructed, the bearing fault impact component is subjected to Hilbert envelope analysis, and the characteristic frequency of the bearing fault is identified.

[0062] As an implementable method, the process of decomposing the original bearing fault signal based on the adaptive noise complete set empirical mode decomposition method to obtain a series of intrinsic mode function components includes:

[0063] The original bearing fault signal is taken as the signal to be processed, and different Gaussian white noises are added to the signal to be processed to obtain a noisy signal set; each noisy signal in the noisy signal set is decomposed based on the empirical mode decomposition method, and the first-order intrinsic mode function component is obtained by ensemble averaging; the first-order intrinsic mode function component is removed from the signal to be processed to obtain a first residual signal; the empirical mode decomposition is continued on the first residual signal after noisy, and the second-order intrinsic mode function component is obtained by ensemble averaging; the second-order intrinsic mode function component is removed from the first residual signal to obtain a second residual signal; and so on, until the obtained residual signal is a monotonic function, the iteration is terminated, and finally a series of intrinsic mode function components are obtained.

[0064] Specifically, the secondary CEEMADN-CCJC noise reduction process includes:

[0065] CEEMDAN adds adaptive white noise to the decomposed signal, which eliminates the modal confusion of EMD decomposition and solves the problem of signal feature damage caused by random noise addition. k (t) is the intrinsic mode function (IMF) component obtained by ensemble average, E j (s) is the jth intrinsic mode component obtained by performing empirical mode decomposition (EMD decomposition) on signal s, ω i (t) is Gaussian white noise that satisfies N(0,1) distribution. Assume that the signal to be processed is y(t), and the decomposition steps of CEEMDAN are as follows:

[0066] (1) Add different Gaussian white noise ω to y(t) i (t), then there is a signal set y i (t) = y(t) + ε0ω i (t), ε0 is the standard deviation of noise. Use EMD to calculate each noisy signal y in the set. i (t) is decomposed and the ensemble average is calculated to obtain the first-order IMF component IMF1(t);

[0067]

[0068] Remove IMF1(t) from the original signal and get the first residual signal r1(t):

[0069] r1(t)=y(t)-IMF1(t) (2)

[0070] (2) For the signal set r1(t)+ε1E1(ω i (t)) is decomposed by EMD and the ensemble average is calculated to obtain the second-order IMF component IMF2(t):

[0071]

[0072] Further get the residual signal r2(t):

[0073] r2(t)=r1(t)-IMF2(t) (4)

[0074] (3) Similarly, for r k (t) = r k-1 (t)-IMF k (t) Add white noise and we get the signal set r k (t)+ε k E k (ω i (t)), perform EMD decomposition on it, and calculate the ensemble average to obtain the k+1th order modal component IMF k+1 (t):

[0075]

[0076] (4) When the residual signal is a monotonic function, the iteration terminates and finally K IMF components are obtained. The final residual signal R(t) is:

[0077]

[0078] That is, the original signal y(t) can be expressed as:

[0079]

[0080] Furthermore, the correlation coefficient can describe the similarity of the periodic characteristics and overall morphology between signals. The closer the amplitude, frequency and morphology are, the larger the correlation coefficient is. For the vibration signal of early rolling bearing fault, because the time-varying impact characteristics that characterize the fault are very weak, the overall appearance will be white noise. When CEEMDAN is used to decompose it, since the carrier frequency of the faulty time-varying impact component is the high-frequency bearing system natural frequency, it will be decomposed into the first few IMF components, and because the morphology is very different from the noise morphology, the corresponding correlation coefficient value is small and stable; and as the high-frequency faulty time-varying impact component is separated, the IMF component dominated by the noise component is further decomposed, and its corresponding correlation coefficient value will increase significantly; further decomposition will be carried out into low-frequency frequency conversion IMF components and small-amplitude residual components, and their correlation coefficient values ​​will gradually decrease. Therefore, the correlation coefficient of the nth-order IMF component is defined as r n, then the correlation coefficient r of each order IMF component is n The correlation coefficient curve formed by the above equation will show a peak shape as a whole; and the correlation coefficient r n+1 Significantly greater than the correlation coefficient r n The steepest position of the peak, that is, the order n corresponding to the maximum jump of the correlation coefficient value, can theoretically be used as the dividing point between the IMF component dominated by the fault time-varying impact component and the IMF component dominated by the noise and frequency conversion component. Based on this, this embodiment proposes a CCJC criterion.

[0081] The Pearson correlation coefficient is used in the CCJC criteria to calculate the correlation coefficient r between vectors:

[0082]

[0083] Where Cov(X,Y) is the covariance of vector X and vector Y, δ X and δ Y Represents the standard deviation of vector X and vector Y respectively.

[0084] Define the maximum jump value p of the IMF component correlation coefficient nmax :

[0085]

[0086] Then it can be determined that component IMFi (1≤i≤n) is an IMF component dominated by fault impulse components, and component IMFi (i>n) is noise, frequency conversion correlation or other residual IMF components.

[0087] The good non-stationary nonlinear signal decomposition ability of CEEMDAN and the CCJC criterion are used to determine the properties of each IMF component. The noise and rotation frequency related components can be fully removed through secondary decomposition and secondary reconstruction to obtain the denoised rolling bearing vibration signal.

[0088] As an implementable approach, the process of ELMSSCT time-varying transient feature enhancement includes:

[0089] ELMSSCT uses CT transformation to expand the signal to TFC space, and by squeezing the CT coefficients to the reference TFC position determined by the local maximum value distributed on the FC plane, it can better concentrate the energy near the time-varying instantaneous frequency of the component signal, thereby improving the time-frequency resolution and the time-frequency domain signal-to-noise ratio. The two theorems and a local maximum FC estimator defined in the specific algorithm of ELMSSCT are the key to determining its analysis effect, as follows:

[0090] The first theorem: Assume a multi-component signal A h (t) and φh (t) are s h (t) is the instantaneous amplitude and instantaneous phase. Define C s (t, μ, c) is the CT transform of s(t) based on the Gaussian window function, where μ is the instantaneous frequency and c is the chirp rate.

[0091] Define a set of non-empty sets Ψ and Ψ h as follows:

[0092]

[0093] If s(t) satisfies the following amplitude conditions:

[0094]

[0095] Where: Λ1=(Δ 2 +1) -1 / 4 , Λ2=(4Δ 2 +1) -1 / 4 , Δ is the separation resolution; α>0. Then Ψ can be expressed as Ψ h The non-intersecting form of is:

[0096]

[0097] The second theorem: Let is the weight s h If the signal s(t) satisfies equation (11), the following results hold:

[0098]

[0099] Where: φ' h (t),φ″ h (t) is the component s h Theoretical values ​​of the instantaneous frequency and chirp rate of (t), A l (t) is the component s l (t), the instantaneous amplitude when l≠h.

[0100] From formula (13), we can see that if T h ≈0, then we have and That is, at this time Can be regarded as the theoretical value φ' h (t),φ″ h (t) is a valid estimate of

[0101] Combining the first and second theorems, we can see that for a multi-component signal that meets the above amplitude and resolution conditions, different components can be located in independent regions in the TFC space, so that components with similar instantaneous frequencies can be effectively distinguished.

[0102] Define the local maximum FC estimator:

[0103]

[0104] Where: Δ μ , Δ c is the frequency resolution and chirp rate resolution. From Theorem 1 and Theorem 2, we can see that the local maximum FC estimator is defined by is the theoretical instantaneous frequency of the component signal φ' h (t) and chirp rate φ″ h An effective estimate of (t) is:

[0105]

[0106] Therefore, by s (t,μ,c) Squeezing, as shown in formula (16), can significantly improve the time-varying instantaneous characteristics of the analyzed signal in the squeezed TFC space The energy concentration on.

[0107]

[0108] Where: γ>0, ensuring C s (t,μ,c)≠0, δ is the Dirac function, η and β are the instantaneous frequency and chirp rate after squeezing.

[0109] Further, By integrating along the chirp rate direction, we can obtain the high-resolution time-frequency spectrum of the signal:

[0110]

[0111] In the formula, is the chirp rate coefficient distribution range.

[0112] In summary, the use of ELMSSCT to process the noise-reduced bearing fault vibration signal can theoretically improve the energy concentration of the fault impact component area in the signal, which is equivalent to enhancing the time-varying instantaneous characteristics of the signal reflecting the fault, and at the same time improve the distinguishability of components with similar frequencies in the signal.

[0113] As an implementable method, a simulation analysis is performed on the method proposed in this embodiment:

[0114] In order to verify the feasibility of the proposed method, the rolling bearing inner ring fault response function M(t) is constructed and superimposed with Gaussian white noise n(t) to obtain the bearing inner ring fault simulation signal Sig(t) as shown in formula (18):

[0115]

[0116] Where: t0 represents the sampling time of a single fault impact, M i (t0) represents the single-cycle impact component caused by the inner race fault, ∑ i M i (t0) represents the i (t0) is a periodic impulse signal repeated with t0 as the time period, A0 is the impulse signal amplitude, C is the system attenuation coefficient, B is the offset, i is the number of impulse repetitions, f r 、f n 、f i They are the bearing rotation frequency, system natural frequency and inner ring fault characteristic frequency respectively.

[0117] Set the bearing system natural frequency f n =4000Hz, bearing rotation frequency f r =20Hz, inner ring fault characteristic frequency f i =170Hz, signal sampling frequency f s =20kHz, sampling points N = 10000, add Gaussian white noise with signal-to-noise ratio SNR = -5dB, and the simulation signal of bearing inner ring fault is as follows Figure 2 As shown, it can be seen that the periodic impact component caused by the fault has been completely submerged by the noise.

[0118] The proposed method is used to Figure 2 The simulation signal of the inner race fault of the bearing is analyzed. First, noise reduction is performed. The first CEEMDAN decomposition obtains 14 IMF components as follows Figure 3 As shown in the figure, it can be seen that the fault impact characteristic component, noise, rotation frequency and its multiple frequency components cannot be accurately distinguished from the time domain waveform of the component; further calculating the correlation coefficient between each IMF component and the original bearing signal, we can get the following: Figure 4 The correlation coefficient curve shown in the figure shows that the correlation coefficient curve is peak-shaped and has an obvious inflection point. According to the proposed CCJC criterion, the steepest position of the curve determines that components IMF1,1 to IMF1,3 are mainly fault impact components, while the remaining components are noise, frequency-related components, and small-amplitude residual components. IMF1,1 to IMF1,3 are superimposed and reconstructed to obtain a reconstructed signal as shown in Figure 5 As shown, it can be seen that the signal-to-noise ratio is improved compared to the original signal.

[0119] The reconstructed signal is decomposed by CEEMDAN twice, and a total of 14 IMF components IMF2,1~IMF2,14 are obtained. The correlation coefficient between each IMF component and the original signal is calculated as follows: Figure 6 As shown in the figure, it can be seen that the correlation coefficient curve still presents a peak shape. According to the CCJC criterion, it can be judged that IMF2,1 is the component dominated by the fault impact component, IMF2,2~IMF2,7 are the main noise components, and IMF2,8~IMF2,14 are low-frequency small-amplitude residual components. Therefore, IMF2,1 is used as the secondary reconstruction signal, and its time domain waveform is as follows Figure 7 As shown in the figure, it can be seen that it presents certain periodic impact characteristics, and the signal-to-noise ratio is further improved; the Hilbert envelope analysis is performed on it, and the results are as follows Figure 7 As shown in the figure, it can be seen that the inner ring fault characteristic frequency f i and its frequency multiples, but there is also interference from the sidebands and residual noise produced by frequency conversion modulation. Figure 8 for Figure 7 The STFT time-frequency diagram of the secondary reconstructed signal is shown. It can be seen that although the signal energy is mainly distributed near the natural frequency of 4000 Hz, due to the complex modulation of the inner race fault characteristic frequency and the rotation frequency on the natural frequency and the influence of residual noise, and the low frequency resolution, the time-varying instantaneous characteristic information reflecting the fault cannot be clearly identified and needs further processing.

[0120] Using ELMSSCT Figure 7 The secondary reconstructed signal shown is subjected to time-varying transient feature enhancement processing. Fig. 9 In order to map the secondary reconstructed signal to the TFC space, the signal time-frequency spectrum is obtained after squeezing and rearranging. It can be seen that the time-frequency energy is highly concentrated in the natural frequency of 4000Hz, and the time-varying instantaneous characteristics are also significantly enhanced. Due to the good resolution of ELMSSCT for components with similar frequencies in the signal, the natural frequency components and nearby related interference components are clearly distinguished. Therefore, only the time-frequency chirp rate coefficients in the natural frequency region are further selected for time domain reconstruction, and the feature-enhanced time domain fault vibration signal is obtained as follows: Fig.10 As shown in Figure 1, it can be seen that the periodic impact characteristics of the fault are very obvious; its Hilbert envelope spectrum is as follows Fig.10 As shown, we can clearly see the characteristic frequency f of the inner race fault i and its doublets, and compared to Figure 7 , the amplitude of the characteristic spectrum line increases, while the frequency modulation sideband and noise are basically eliminated, which proves that the proposed method can effectively reduce noise and enhance fault characteristics. In particular, the elimination of the frequency modulation sideband interference component makes this method suitable for fault feature extraction when the structure or working conditions are complex and the sidebands interfere with each other.

[0121] As a specific example, the method proposed in this example is experimentally analyzed:

[0122] Rolling bearing fault test experiments were carried out on a rotating machinery fault test bench. The data acquisition equipment was a 40-channel data acquisition instrument from LMS, and acceleration data was picked up through a PCB three-way vibration acceleration sensor. The experimental bearing was an SKF-6311 deep groove ball bearing, installed on the non-driving end of the gearbox input shaft. Laser wire cutting was used to cut grooves on the inner and outer rings to simulate local faults in the rolling bearing raceway. The sampling frequency for the inner and outer ring fault experiments was 8192Hz, the input shaft speed was 680r / min, and the theoretical outer ring fault characteristic frequency was calculated to be 34.3Hz and the theoretical inner ring fault characteristic frequency was 56.7Hz.

[0123] The outer ring fault vibration signal is collected. The time domain waveform of the signal has no obvious periodic impact characteristics and shows white noise characteristics. The proposed method is used to analyze it. First, the original signal is denoised by secondary CEEMDAN-CCJC to obtain the denoised signal and its Hilbert envelope spectrum. The signal-to-noise ratio of the denoised signal is significantly improved compared with the original signal, showing the periodic impact characteristics caused by the fault; there is an outer ring fault characteristic frequency f in the envelope spectrum o The noise reduction effect is certain, but it can also be seen that there is still a certain amount of noise energy distributed in the entire frequency band, which affects the recognition effect of the fault characteristic frequency spectrum. Based on the STFT time-frequency diagram of the signal after noise reduction, it can be seen that the energy is concentrated in the two frequency bands of 2500-2700Hz and 3500-3700Hz, corresponding to the two inherent frequency bands of the outer ring fault bearing, but due to the complex modulation and residual noise interference, the time-varying instantaneous characteristics in the two inherent frequency bands are not clear and cannot be accurately identified.

[0124] ELMSSCT is further used to enhance the time-varying instantaneous features of the denoised signal. The energy of the processed signal is highly concentrated in the second-order natural frequencies of the outer ring fault bearing, especially the time-varying instantaneous features of the fault in the first-order natural frequency region become very clear. Therefore, the bearing fault signal is reconstructed using the time-frequency chirp rate coefficient near the first-order natural frequency region. The signal after feature enhancement has obvious periodic fault impact characteristics; a very clear outer ring fault characteristic frequency f can be observed from the Hilbert envelope spectrum. o The white noise originally distributed in the entire frequency band is greatly reduced, which fully demonstrates the effectiveness of the proposed method.

[0125] The experimental signal of the bearing inner ring fault is collected. The periodic impact component representing the fault in the signal is submerged in the noise and difficult to identify. The proposed method is also used to reduce noise, enhance and extract time-varying instantaneous features, and the steps are consistent with the above. The denoised signal and its Hilbert envelope spectrum and STFT time-frequency diagram are obtained, as well as the TFC time-frequency diagram after ELMSSCT enhancement of the time-varying instantaneous features, the reconstructed bearing signal and its Hilbert envelope spectrum. It can be seen that after the processing of the proposed method, the interference components such as noise and rotation frequency in the original bearing vibration signal are basically eliminated, while the time-varying instantaneous features reflecting the bearing fault are enhanced and can be clearly identified and extracted from the Hilbert envelope spectrum. The above analysis fully illustrates the good effect of the proposed method in processing the measured bearing vibration signal.

[0126] This embodiment clearly extracts the characteristic frequency information reflecting the bearing fault by analyzing the rolling bearing fault simulation signal and the experimental signal, and verifies the effectiveness of the proposed method in reducing noise interference and enhancing the time-varying transient characteristics of the fault. It is especially suitable for early fault signals of bearings with severe noise and complex time-frequency characteristics.

[0127] Embodiment 2

[0128] This embodiment further provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method.

[0129] Embodiment 3

[0130] This embodiment also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the method are implemented.

[0131] Embodiment 4

[0132] This embodiment also provides a computer program product, including a computer program, which implements the steps of the method when executed by a processor.

[0133] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for enhancing and extracting time-varying transient features of rolling bearing faults, characterized in that: The following steps are involved: Collect original bearing fault signals; Decomposing the original bearing fault signal based on an adaptive noise complete set empirical mode decomposition method to obtain a series of intrinsic mode function components; The properties of the intrinsic mode function components are distinguished based on the correlation coefficient jump criterion, and denoising is performed by a secondary decomposition and secondary reconstruction method to obtain a bearing fault signal after denoising; Based on the local maximum frequency chirp rate synchronous compression chirp transform method, the de-noised bearing fault signal is mapped to the time-frequency chirp rate space through chirp transform; The time-frequency chirp rate coefficient reflecting the time-varying instantaneous characteristics of the fault is extracted from the squeezed time-frequency chirp rate space, the bearing fault impact component is reconstructed, the bearing fault impact component is subjected to Hilbert envelope analysis, and the characteristic frequency of the bearing fault is identified.

2. The method according to claim 1, characterized in that The process of decomposing the original bearing fault signal based on the adaptive noise complete set empirical mode decomposition method to obtain a series of intrinsic mode function components includes: The original bearing fault signal is taken as the signal to be processed, and different Gaussian white noises are added to the signal to be processed to obtain a noisy signal set; each noisy signal in the noisy signal set is decomposed based on the empirical mode decomposition method, and the first-order intrinsic mode function component is obtained by ensemble averaging; the first-order intrinsic mode function component is removed from the signal to be processed to obtain a first residual signal; the empirical mode decomposition is continued on the first residual signal after noisy, and the second-order intrinsic mode function component is obtained by ensemble averaging; the second-order intrinsic mode function component is removed from the first residual signal to obtain a second residual signal; and so on, until the obtained residual signal is a monotonic function, the iteration is terminated, and finally a series of intrinsic mode function components are obtained.

3. The method according to claim 1, characterized in that The process of distinguishing the properties of the intrinsic mode function components based on the correlation coefficient jump criterion includes: The correlation coefficient jump criterion adopts the Pearson correlation coefficient to obtain the correlation coefficient value between the vectors, obtains the maximum jump value of the correlation coefficient of the intrinsic mode function component based on the correlation coefficient value between the vectors, determines the maximum jump value of the correlation coefficient of the intrinsic mode function component, and distinguishes the noise component and the non-noise component in the intrinsic mode function component.

4. The method according to claim 1, characterized in that: The process of mapping the de-noised bearing fault signal to the time-frequency chirp rate space through chirp transform based on the local maximum frequency chirp rate synchronous compression chirp transform method includes: The first theorem, the second theorem and the local maximum FC estimator are constructed. Based on the first theorem, the second theorem and the local maximum FC estimator, the denoised bearing fault signal is mapped to an independent region of the time-frequency chirp rate space through chirp transform.

5. The method according to claim 4, characterized in that The formula of the first theorem is as follows: Define a set of non-empty sets Ψ and Ψ h as follows: in, A h( t ) and φ h (t) are s h (t) is the instantaneous amplitude and instantaneous phase, C s( t,μ,c ) is the CT transform of s(t) based on Gaussian window function, μ is the instantaneous frequency, and c is the chirp rate; When s(t) satisfies the following amplitude conditions: In the formula, Λ1=(Δ 2 +1) -1 / 4 , Λ2=(4Δ 2 +1) -1 / 4 , Δ is the separation resolution; α>0; Ψ is represented by Ψ h The non-intersecting form of is:

6. The method according to claim 5, characterized in that The second theorem is expressed as follows: When s(t) satisfies the amplitude condition described in claim 5, the following results hold: In the formula, is the weight s h (t) is an estimate of the instantaneous frequency and chirp rate; φ' h( t ) 、φ' h ' ( t ) The components s h Theoretical values ​​of the instantaneous frequency and chirp rate of (t), A l( t ) is the weight s l( t ) , the instantaneous amplitude of l≠h.

7. The method according to claim 6, characterized in that The formula of the local maximum FC estimator is as follows: In the formula, Δ μ , Δ c They are frequency resolution and chirp rate resolution respectively.

8. A computer device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images